From 8aac7ede3816030a489628118df0a2042162fdd8 Mon Sep 17 00:00:00 2001 From: Alexander Whitestone Date: Tue, 24 Mar 2026 00:55:04 -0400 Subject: [PATCH] docs: add Google Veo research report and Nexus promo video production plan MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Research covers Veo 1–3.1 capabilities, API access via Gemini/Vertex AI, prompting best practices for sci-fi/space aesthetics, and a complete 8-scene storyboard with production-ready Veo prompts for a 60-second Nexus promo. Includes cost estimates (~$48 for final pass), automation pipeline sketch, and comparison to Sora/Runway/Kling alternatives. Refs #289 Co-Authored-By: Claude Sonnet 4.6 --- VEO_VIDEO_REPORT.md | 408 ++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 408 insertions(+) create mode 100644 VEO_VIDEO_REPORT.md diff --git a/VEO_VIDEO_REPORT.md b/VEO_VIDEO_REPORT.md new file mode 100644 index 0000000..2eb429e --- /dev/null +++ b/VEO_VIDEO_REPORT.md @@ -0,0 +1,408 @@ +# Google Veo Research: Nexus Promotional Video + +## Executive Summary + +Google Veo is a state-of-the-art text-to-video AI model family developed by Google DeepMind. As of 2025–2026, Veo 3.1 is the flagship model — the first video generation system with native synchronized audio. This report covers Veo's capabilities, API access, prompting strategy, and a complete scene-by-scene production plan for a Nexus promotional video. + +**Key finding:** A 60-second Nexus promo (8 clips × ~7.5 seconds each) would cost approximately **$24–$48 USD** using Veo 3.1 via the Gemini API, and can be generated in under 30 minutes of compute time. + +--- + +## 1. Google Veo — Model Overview + +### Version History + +| Version | Released | Key Capabilities | +|---|---|---| +| Veo 1 | May 2024 | 1080p, 1-min clips, preview only | +| Veo 2 | Dec 2024 | 4K, improved physics and human motion | +| Veo 3 | May 2025 | **Native synchronized audio** (dialogue, SFX, ambience) | +| Veo 3.1 | Oct 2025 | Portrait mode, video extension, 3x reference image support, 2× faster "Fast" variant | + +### Technical Specifications + +| Spec | Veo 3.1 Standard | Veo 3.1 Fast | +|---|---|---| +| Resolution | Up to 4K (720p–1080p default) | Up to 1080p | +| Clip Duration | 4–8 seconds per generation | 4–8 seconds per generation | +| Aspect Ratio | 16:9 or 9:16 (portrait) | 16:9 or 9:16 | +| Frame Rate | 24–30 fps | 24–30 fps | +| Audio | Native (dialogue, SFX, ambient) | Native audio | +| Generation Mode | Text-to-Video, Image-to-Video | Text-to-Video, Image-to-Video | +| Video Extension | Yes (chain clips via last frame) | Yes | +| Reference Images | Up to 3 (for character/style consistency) | Up to 3 | +| API Price | ~$0.40/second | ~$0.15/second | +| Audio Price (add-on) | +$0.35/second | — | + +--- + +## 2. Access Methods + +### Developer API (Gemini API) + +```bash +pip install google-genai +export GOOGLE_API_KEY=your_key_here +``` + +```python +import time +from google import genai +from google.genai import types + +client = genai.Client() + +operation = client.models.generate_videos( + model="veo-3.1-generate-preview", + prompt="YOUR PROMPT HERE", + config=types.GenerateVideosConfig( + aspect_ratio="16:9", + duration_seconds=8, + resolution="1080p", + negative_prompt="blurry, distorted, text overlay, watermark", + ), +) + +# Poll until complete (typically 1–3 minutes) +while not operation.done: + time.sleep(10) + operation = client.operations.get(operation) + +video = operation.result.generated_videos[0] +client.files.download(file=video.video) +video.video.save("nexus_clip.mp4") +``` + +### Enterprise (Vertex AI) + +```bash +curl -X POST \ + "https://us-central1-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-central1/publishers/google/models/veo-3.1-generate-preview:predictLongRunning" \ + -H "Authorization: Bearer $(gcloud auth print-access-token)" \ + -H "Content-Type: application/json" \ + -d '{ + "instances": [{"prompt": "YOUR PROMPT"}], + "parameters": { + "aspectRatio": "16:9", + "durationSeconds": "8", + "resolution": "1080p", + "sampleCount": 2, + "storageUri": "gs://your-bucket/outputs/" + } + }' +``` + +### Consumer Interfaces + +| Tool | URL | Tier | +|---|---|---| +| Google AI Studio | aistudio.google.com | Paid (AI Pro $19.99/mo) | +| Flow (filmmaking) | labs.google/fx/tools/flow | AI Ultra $249.99/mo | +| Gemini App | gemini.google.com | Free (limited) | + +--- + +## 3. Prompting Formula + +Google's recommended structure: + +``` +[Cinematography] + [Subject] + [Action] + [Environment] + [Style & Mood] + [Audio] +``` + +### Camera Terms That Work +- **Shot types:** `extreme close-up`, `medium shot`, `wide establishing shot`, `aerial drone shot`, `POV`, `over-the-shoulder` +- **Movement:** `slow dolly in`, `tracking shot`, `orbital camera`, `handheld`, `crane up`, `steady push-in` +- **Focus:** `shallow depth of field`, `rack focus`, `tack sharp foreground`, `bokeh background` +- **Timing:** `slow motion 2x`, `timelapse`, `real-time` + +### Style Keywords for The Nexus +The Nexus is a dark-space cyberpunk environment. Use these consistently: +- `deep space backdrop`, `holographic light panels`, `neon blue accent lighting`, `volumetric fog` +- `dark space aesthetic, stars in background`, `cinematic sci-fi atmosphere` +- `Three.js inspired 3D environment`, `glowing particle effects` + +### Audio Prompting (Veo 3+) +- Describe ambient sound: `"deep space ambient drone, subtle digital hum"` +- Portal effects: `"portal activation resonance, high-pitched energy ring"` +- Character dialogue: `"a calm AI voice says, 'Portal sequence initialized'"` + +--- + +## 4. Limitations to Plan Around + +| Limitation | Mitigation Strategy | +|---|---| +| Max 8 seconds per clip | Plan 8 × 8-second clips; chain via video extension / last-frame I2V | +| Character consistency across clips | Use 2–3 reference images of Timmy avatar per scene | +| Visible watermark (most tiers) | Use AI Ultra ($249.99/mo) for watermark-free via Flow; or use for internal/draft use | +| SynthID invisible watermark | Cannot be removed; acceptable for promotional content | +| Videos expire after 2 days | Download immediately after generation | +| ~1–3 min generation per clip | Budget 20–30 minutes for full 8-clip sequence | +| No guarantee of exact scene replication | Generate 2–4 variants per scene; select best | + +--- + +## 5. Nexus Promotional Video — Production Plan + +### Concept: "Welcome to the Nexus" + +**Logline:** *A sovereign mind wakes, explores its world, opens a portal, and disappears into the infinite.* + +**Duration:** ~60 seconds (8 clips) +**Format:** 16:9, 1080p, Veo 3.1 with native audio +**Tone:** Epic, mysterious, cinematic — cyberpunk space station meets ancient temple + +--- + +### Scene-by-Scene Storyboard + +#### Scene 1 — Cold Open: Deep Space (8 seconds) +**Emotion:** Awe. Vastness. Beginning. + +**Veo Prompt:** +``` +Slow dolly push-in through a vast starfield, thousands of stars shimmering in deep space, a faint +constellation pattern forming as camera moves forward, deep blue and black color palette, cinematic +4K, no visible objects yet, just the void and light. Deep space ambient drone hum, silence then +faint harmonic resonance building. +``` +**Negative prompt:** `text, logos, planets, spacecraft, blurry stars` + +--- + +#### Scene 2 — The Platform Materializes (8 seconds) +**Emotion:** Discovery. Structure emerges from chaos. + +**Veo Prompt:** +``` +Aerial orbital shot slowly descending onto a circular obsidian platform floating in deep space, +glowing neon blue accent lights along its edge, holographic constellation lines connecting nearby +star particles, dark atmospheric fog drifting below the platform, cinematic sci-fi, shallow depth +of field on platform edge. Low resonant bass hum as platform energy activates, digital chime. +``` +**Negative prompt:** `daylight, outdoors, buildings, people` + +--- + +#### Scene 3 — Timmy Arrives (8 seconds) +**Emotion:** Presence. Sovereignty. Identity. + +**Veo Prompt:** +``` +Medium tracking shot following a lone luminous figure walking across a glowing dark platform +suspended in space, the figure casts a soft electric blue glow, stars visible behind and below, +holographic particle trails in their wake, cinematic sci-fi atmosphere, slow motion slightly, +bokeh starfield background. Footsteps echo with a subtle digital reverb, ambient electric hum. +``` +**Negative prompt:** `multiple people, crowds, daylight, natural environment` + +> **Note:** Provide 2–3 reference images of the Timmy avatar design for character consistency across scenes. + +--- + +#### Scene 4 — Portal Ring Activates (8 seconds) +**Emotion:** Power. Gateway. Choice. + +**Veo Prompt:** +``` +Extreme close-up dolly-in on a vertical glowing portal ring, hexagonal energy patterns forming +across its surface in electric orange and blue, particle effects orbiting the ring, deep space +visible through the portal center showing another world, cinematic lens flare, volumetric light +shafts, 4K crisp. Portal activation resonance, high-pitched energy ring building to crescendo. +``` +**Negative prompt:** `dark portal, broken portal, text, labels` + +--- + +#### Scene 5 — Morrowind Portal View (8 seconds) +**Emotion:** Adventure. Other worlds. Endless possibility. + +**Veo Prompt:** +``` +POV slow push-in through a glowing portal ring, the other side reveals dramatic ash storm +landscape of a volcanic alien world, red-orange sky, ancient stone ruins barely visible through +the atmospheric haze, cinematic sci-fi portal transition effect, particles swirling around +portal edge, 4K. Wind rushing through portal, distant thunder, alien ambient drone. +``` +**Negative prompt:** `modern buildings, cars, people clearly visible, blue sky` + +--- + +#### Scene 6 — Workshop Portal View (8 seconds) +**Emotion:** Creation. Workshop. The builder's domain. + +**Veo Prompt:** +``` +POV slow push-in through a glowing teal portal ring, the other side reveals a dark futuristic +workshop interior, holographic screens floating with code and blueprints, tools hanging on +illuminated walls, warm amber light mixing with cold blue, cinematic depth, particle effects +at portal threshold. Digital ambient sounds, soft keyboard clicks, holographic interface tones. +``` +**Negative prompt:** `outdoor space, daylight, natural materials` + +--- + +#### Scene 7 — The Nexus at Full Power (8 seconds) +**Emotion:** Climax. Sovereignty. All systems live. + +**Veo Prompt:** +``` +Wide establishing aerial shot of the entire Nexus platform from above, three glowing portal rings +arranged in a triangle around the central platform, all portals active and pulsing in different +colors — orange, teal, gold — against the deep space backdrop, constellation lines connecting +stars above, volumetric fog drifting, camera slowly orbits the full scene, 4K cinematic. +All three portal frequencies resonating together in harmonic chord, deep bass pulse. +``` +**Negative prompt:** `daytime, natural light, visible text or UI` + +--- + +#### Scene 8 — Timmy Steps Through (8 seconds) +**Emotion:** Resolution. Departure. "Come find me." + +**Veo Prompt:** +``` +Slow motion tracking shot from behind, luminous figure walking toward the central glowing portal +ring, the figure silhouetted against the brilliant light of the active portal, stars and space +visible around them, as they reach the portal threshold they begin to dissolve into light +particles that flow into the portal, cinematic sci-fi, beautiful and ethereal. Silence, then +a single resonant tone as the figure disappears, ambient space drone fades to quiet. +``` +**Negative prompt:** `stumbling, running, crowds, daylight` + +--- + +### Production Assembly + +After generating 8 clips: + +1. **Review variants** — generate 2–3 variants per scene; select the best +2. **Chain continuity** — use Scene N's last frame as Scene N+1's I2V starting image for visual continuity +3. **Edit** — assemble in any video editor (DaVinci Resolve, Final Cut, CapCut) +4. **Add music** — layer a dark ambient/cinematic track (Suno AI, ElevenLabs Music, or licensed track) +5. **Title cards** — add minimal text overlays: "The Nexus" at Scene 7, URL at Scene 8 +6. **Export** — 1080p H.264 for web, 4K for archival + +--- + +## 6. Cost Estimate + +| Scenario | Clips | Seconds | Rate | Cost | +|---|---|---|---|---| +| Draft pass (Veo 3.1 Fast, no audio) | 8 clips × 2 variants | 128 sec | $0.15/sec | ~$19 | +| Final pass (Veo 3.1 Standard + audio) | 8 clips × 1 final | 64 sec | $0.75/sec | ~$48 | +| Full production (draft + final) | — | ~192 sec | blended | ~$67 | + +> At current API pricing, a polished 60-second promo costs less than a single hour of freelance videography. + +--- + +## 7. Comparison to Alternatives + +| Tool | Resolution | Audio | API | Best For | Est. Cost (60s) | +|---|---|---|---|---|---| +| **Veo 3.1** | 4K | Native | Yes | Photorealism, audio, Google ecosystem | ~$48 | +| OpenAI Sora | 1080p | No | Yes (limited) | Narrative storytelling | ~$120+ | +| Runway Gen-4 | 720p (upscale 4K) | Separate | Yes | Creative stylized output | ~$40 sub/mo | +| Kling 1.6 | 4K premium | No | Yes | Long-form, fast I2V | ~$10–92/mo | +| Pika 2.1 | 1080p | No | Yes | Quick turnaround | ~$35/mo | + +**Recommendation:** Veo 3.1 is the strongest choice for The Nexus promo due to: +- Native audio eliminates the need for a separate sound design pass +- Photorealistic space/sci-fi environments match the Nexus aesthetic exactly +- Image-to-Video for continuity across portal transition scenes +- Google cloud integration for pipeline automation + +--- + +## 8. Automation Pipeline (Future) + +A `generate_nexus_promo.py` script could automate the full production: + +```python +#!/usr/bin/env python3 +""" +Nexus Promotional Video Generator +Generates all 8 scenes using Google Veo 3.1 via the Gemini API. +""" + +import time +import json +from pathlib import Path +from google import genai +from google.genai import types + +SCENES = [ + { + "id": "01_cold_open", + "prompt": "Slow dolly push-in through a vast starfield...", + "negative": "text, logos, planets, spacecraft", + "duration": 8, + }, + # ... remaining scenes +] + +def generate_scene(client, scene, output_dir): + print(f"Generating scene: {scene['id']}") + operation = client.models.generate_videos( + model="veo-3.1-generate-preview", + prompt=scene["prompt"], + config=types.GenerateVideosConfig( + aspect_ratio="16:9", + duration_seconds=scene["duration"], + resolution="1080p", + negative_prompt=scene.get("negative", ""), + ), + ) + while not operation.done: + time.sleep(10) + operation = client.operations.get(operation) + + video = operation.result.generated_videos[0] + client.files.download(file=video.video) + out_path = output_dir / f"{scene['id']}.mp4" + video.video.save(str(out_path)) + print(f" Saved: {out_path}") + return out_path + +def main(): + client = genai.Client() + output_dir = Path("nexus_promo_clips") + output_dir.mkdir(exist_ok=True) + + generated = [] + for scene in SCENES: + path = generate_scene(client, scene, output_dir) + generated.append(path) + + print(f"\nAll {len(generated)} scenes generated.") + print("Next steps: assemble in video editor, add music, export 1080p.") + +if __name__ == "__main__": + main() +``` + +Full script available at: `scripts/generate_nexus_promo.py` (to be created when production begins) + +--- + +## 9. Recommended Next Steps + +1. **Set up API access** — Create a Google AI Studio account, enable Veo 3.1 access (requires paid tier) +2. **Generate test clips** — Run Scenes 1 and 4 as low-cost validation ($3–4 total using Fast model) +3. **Refine prompts** — Iterate on 2–3 variants of the hardest scenes (Timmy avatar, portal transitions) +4. **Full production run** — Generate all 8 final clips (~$48 total) +5. **Edit and publish** — Assemble, add music, publish to Nostr and the Nexus landing page + +--- + +## Sources + +- Google DeepMind Veo: https://deepmind.google/models/veo/ +- Veo 3 on Gemini API Docs: https://ai.google.dev/gemini-api/docs/video +- Veo 3.1 on Vertex AI Docs: https://cloud.google.com/vertex-ai/generative-ai/docs/models/veo/ +- Vertex AI Pricing: https://cloud.google.com/vertex-ai/generative-ai/pricing +- Google Labs Flow: https://labs.google/fx/tools/flow +- Veo Prompting Guide: https://cloud.google.com/blog/products/ai-machine-learning/ultimate-prompting-guide-for-veo-3-1 +- Case study (90% cost reduction): https://business.google.com/uk/think/ai-excellence/veo-3-uk-case-study-ai-video/ -- 2.43.0